Advanced State Estimation and Thermal Fault Detection for Hybrid-electric Aircraft Using Neural Kalman Filtering and Model Predictive Control

Hybrid-electric airplanes (HEA) need very smart and reliable thermal management systems that can deal with the tough conditions of power, weight and safety.This research proposes a unique combination of a spiking neural Kalman filter (SNN-KF) and a neural model predictive controller (NN-MPC) that has been experimentally verified for real-time, certifiable thermal control on neuromorphic hardware.The hardware-in-the-loop (HIL) tests done on the Intel Loihi 2 neuromorphic processor show a power consumption of 22.7 mW, worst-case latency of 34 ms, 0.082 K temperature RMSE, and no constraint violations during gradual thermal faults and ambient disturbances.The system effectively adjusts to battery-coolant heat transfer deterioration and chiller efficiency faults, which was confirmed through 20 Monte Carlo simulations with ±15% model mismatch-while saving 11.8% cooling energy as compared to the conventional MPC.Hence, these findings prove that the use of neuromorphic control can satisfy the real-time avionics requirements (DO-178C/ASIL-D) at a milliwatt power level, thereby allowing the intelligence of electric aircraft, satellites, and UAVs of the next generation to be deployed.